Table Of ContentThermal territories of the abdomen after Caesarean
Section birth : infrared thermography and analysis
CHILDS, Charmaine <http://orcid.org/0000-0002-1558-5633>, SIRAJ,
Mahbubur Rob, FAIR, Frankie <http://orcid.org/orcid.org/0000-0001-7613-
3393>, SELVAN, Arul <http://orcid.org/0000-0001-9222-5538>, SOLTANI,
Hora <http://orcid.org/0000-0001-9611-6777>, WILMOTT, Jon and FARRELL,
Rom
Available from Sheffield Hallam University Research Archive (SHURA) at:
http://shura.shu.ac.uk/12563/
This document is the author deposited version. You are advised to consult the
publisher's version if you wish to cite from it.
Published version
CHILDS, Charmaine, SIRAJ, Mahbubur Rob, FAIR, Frankie, SELVAN, Arul,
SOLTANI, Hora, WILMOTT, Jon and FARRELL, Rom (2016). Thermal territories of
the abdomen after Caesarean Section birth : infrared thermography and analysis.
Journal of wound care, 25 (9), 499-512.
Copyright and re-use policy
See http://shura.shu.ac.uk/information.html
Sheffield Hallam University Research Archive
http://shura.shu.ac.uk
Thermal territories of the abdomen after caesarean section birth: infrared
thermography and analysis approaches to surgical site assessment
Professor Charmaine Childs , Professor of Clinical Science, Centre for Health and Social
Care Research, Sheffield Hallam University, Montgomery House, 32 Collegiate Crescent,
Sheffield S102BP
Dr Mahbubur Rob Siraj, ST7 Obstetrics & Gynaecology, Jessop Wing, Sheffield Teaching
Hospital NHS Trust, Tree Root Walk, Sheffield, S10 2SF
Ms Frankie J Fair, Midwifery Researcher, Centre for Health and Social Care Research,
Sheffield Hallam University, Montgomery House, 32 Collegiate Crescent
Sheffield, S10 2BP
Dr Arul N Selvan, Associate Lecturer, Materials and Engineering Research Institute,
Sheffield Hallam University, Howard Street, Sheffield, S1 1WB
Hora Soltani, Professor of Maternal and Infant Health, Centre for Health and Social Care
Research, Sheffield Hallam University, 32 Collegiate Crescent, Sheffield, S10 2BP
Dr Jon Wilmott, EPSRC Research Fellow, University of Sheffield, Portobello Centre,
Sheffield S1 3JD
Mr Tom Farrell, Consultant, Department of Obstetrics and Gynaecology, Jessop Wing,
Sheffield Teaching Hospital NHS Trust Hospital, Tree Root Walk, Sheffield S102SF
*Corresponding author
Professor Charmaine Childs
[email protected]
1
ABSTRACT
Objective: To develop and refine qualitative mapping and quantitative analysis techniques to
define 'thermal territories' of the post-partum abdomen, the caesarean section site and the
infected surgical wound. In addition, to explore women's perspectives on thermal imaging
and acceptability as a method for infection screening.
Method: Prospective feasibility study undertaken at a large University teaching hospital,
Sheffield UK. Infrared thermal imaging of the abdomen was undertaken at the bedside on
the first two days after elective caesarean section. Target recruitment: six women in each of
three body mass index (BMI) categories (normal, 18.5 to 24.9kg/m²; overweight 25 to
<30kg/m²; obese ≥30kg/m²). Additionally, women presenting to the ward with wound
infection were eligible for inclusion in the study. Perspectives on the use of thermal imaging
and its practicality were also explored via semi-structured interviews and analysed using
thematic content analysis.
Results: Twenty women were recruited. All had undergone caesarean section. From the
booking BMI, eight women were obese (including two women with infected wounds), six
women were overweight and six women had a normal BMI. Temperature (oC) profiling and
pixel clustering segmentation (Hierarchical Clustering Segmentation, HCS) revealed
characteristic features of thermal territories between scar and adjacent regions. Differences
in scar thermal intensity profiles exist between healthy scar and infected wounds; features
that have potential for wound surveillance. Maximum temperature differences (deltaT)
between healthy skin reference and wound site, exceed 2oC in women with established
wound infection. At day 2, two women had a scar thermogram with features observed in the
‘’infected’’ wound thermogram.
Thermal imaging at early and later times after caesarean birth is feasible and acceptable.
Women reported potential benefits of the technique for future wound infection screening.
Conclusion: Thermal intensity profiling and HCS for pixel cluster dissimilarity between scar
and adjacent healthy skin has potential as a method for the development of techniques
targeted to early infection surveillance in women after caesarean section.
Key words;
Thermal imaging, infrared thermography, abdomen, surgical site infection, Caesarean
section, infection surveillance.
2
INTRODUCTION
Complications and adverse events occur during and after surgery.1 Of the post-operative
complications, wound infection is common especially after abdominal surgery.2 Surgical site
infection (SSI) accounts for 21.8% of healthcare associated infections in the US.3 In the UK,
the Health Protection Agency 4 cites SSI as the third most frequent healthcare-associated
infection. The consequences being increased morbidity, a longer stay in hospital, greater
antibiotic use and increased NHS costs.5 Vulnerable patients are those with serious co-
morbidity; the elderly 6 and obese patients. 7,8 Being overweight or obese and undergoing
surgery is known to be an independent risk factor for SSI particularly in colorectal
surgery.9,10 Obesity and SSI risk is now a matter for increasing concern in women
undergoing caesarean section.11,12
After abdominal delivery and caesarean section, infection can occur in all three of the SSI
categories; superficial incisional SSI, deep incisional SSI and organ (or space) SSI. 13
Typically infection of skin and superficial tissues is most common after caesarean birth.
Given the reported risk associated with raised body mass index (BMI)14 there is concern that
overweight and obese women who deliver by caesarean section are most vulnerable to
developing SSI.15,16
We have shown previously that thermography has potential as a method for infection
surveillance after colorectal surgery 17 where SSI rates range between 19 and 32%.18 Within
operational distances of 1 metre from the target field of view (FOV) optimal focus of digital
and infrared images over the abdomen and across a range of BMI reveal qualitative thermal
mapping characteristics which support thermal imaging as a credible tool for infection
surveillance.17 In the current study our aims were to further explore the potential for
thermography in women undergoing caesarean section, specifically to (i) define the 'thermal
territories' of the post-partum abdomen, the surgical site, and the infected surgical wound (ii)
develop robust qualitative mapping and quantitative analysis systems (iii) seek the
perspectives of women about the potential for thermal imaging as a future wound
surveillance technique.
METHODS
Study Design
A prospective feasibility and exploratory study was undertaken over a period of seven
months. The objective was to recruit a sample of six women with a booking BMI across each
of three categories; normal (NORM 18.5 to 24.9kg/m²) overweight (OV 25 to <30kg/m²),
obese (OB ≥30kg/m²) and who gave birth by elective caesarean section and who would
remain as a hospital admission for a minimum of two days after the birth.
In the event that a woman presented with an infected wound (irrespective of booking BMI)
and were admitted to the postpartum ward during the study period, the aim was to recruit
these women to form an ''infected wound" group of subjects. Local National Health Service
(NHS) ethics approval was obtained before recruitment and data collection commenced.
Participants and Procedures
Based on the booking BMI, eligible women were identified by a member of the study team
before the birth and invited to participate after their caesarean section delivery. Interested
3
women were given a participant information sheet. Once the participant had agreed, a
consent form was provided with an explanation of the study objectives and procedures and
with opportunity for women to discuss and ask questions about any aspects of the study. In
this study, lower abdominal, transverse incision was performed in all women of the method
described by Pfannenstiel.19 It is preferred for its cosmetic advantage, with the curve of the
incision in a natural fold of skin.
Once the baby was born, women were approached once more to ensure that they were
comfortable to continue to participate in the study.
Study information and participant characteristics (early pregnancy height and weight, age,
parity, surgical procedure, drugs administered, body temperature, socio-economic indicators
(e.g. postcode) were collected from maternal records.
Temperature measurements
Ambient conditions and body temperature (oC) measurements were made at the bedside in
either a single room or multi-occupancy ward (maximum four beds). Ambient conditions for
room temperature, (oC) relative humidity (% RH) and air velocity (m.sec-1) were measured
(Kestrel 3000; Weather Meter; Richard Paul Russell Ltd, Hampshire UK).
Body temperature was measured using infrared thermometry ''scanning'' of the skin
overlying a part of the course of the temporal artery 20 (temporal artery thermometer,
Temporal Scanner, Exergen, Watertown, MA, USA). Information of drugs likely to affect the
pattern of body temperature was obtained from the drug chart.
Infrared thermography and imaging protocol
Thermal imaging was undertaken using a portable thermal camera (FLIR Systems T450sc)
with IR detector pixel resolution of 320x240 (Thermal Vision Research Ltd, Somerset, UK).
The camera was first calibrated and certified using a blackbody source (FLIR Systems Inc,
Wilsonville, USA). Thermal sensitivity, <30mK with accuracy ±1oC for ambient temperature
within 15-35oC and with spatial resolution 1.36mRad, image frequency 60mH. Reflective
ambient radiation was obtained via imaging of reflective material.
The abdomen was exposed and any covers or dressings removed and with a duration of 10
min allowed for skin temperature to stabilise. The area from the umbilicus to ischial crest and
within the region of the surgical scar was exposed and the camera focussed to this region of
interest (ROI). A thin cotton sheet was placed over the suprapubic region. Where structures
(hair, suture material, suture beads) were visible, these structures served to aid image focus.
Three to six consecutive images were taken to obtain best possible image composition and
clarity of focus, appearing on the camera screen in ''iron'' colour palette and using
‘’automatic’’ setting.
Image processing and analysis
A hierarchy of approaches were taken to the eventual development of a ''bespoke'' analysis
process for this application; a) FLIR proprietary analysis tools of the of ROI
(qualitative/quantitative) b) edge detection algorithm for wound scar analysis c) Hierarchical
4
Clustering-based Segmentation (HCS) 21 of regions with distinct variability across wound
and adjacent skin territories.
i) Image processing
Data was downloaded from the IR camera to a stand-alone PC. Participant identifiers were
removed from all data (demographic, images).
Post-processing was undertaken at the end of each day. FLIR systems software analysis
parameters were adjusted to allow ''apparent temperature'' to be converted to “absolute’’
temperature (oC). Here FLIR Research IR proprietary software allows adjustment to set
parameters for; emissivity (set to 0.98) distance (set to 60cm, i.e. distance from camera to
abdomen) relative humidity (%RH) and reflective ambient temperature (obtained from the
temperature of reflective material before measurements commenced). Once parameter
adjustments were made, a colour palette for pixel representation of temperature was
selected. Rainbow high contrast (Rainbow HC) colour palette was selected and used
throughout. In addition to the temperature (oC) map, ''raw'' radiometric data was produced for
each participants set of thermal images. Using Rainbow HC, the convention for the colour
palette is for higher temperature/higher radiation intensity to be ''brightest'' and lower
temperatures/lower radiation intensity to be represented by ''darker'' colours. From the
bedside image FOV, the visual image was cropped to an image of the lower abdomen only
(image not shown) and from the corresponding thermogram a rectangular box (standardised
to 268x81 pixels) was used to identify the ROI. This incorporated the scar and immediate
surrounding skin.
From each pixel within the ROI, two sets of data were obtained: temperature (oC) (Fig 1a)
and radiometric values (Fig 1b). The scale for temperature was adjusted and standardised
for all participants (range, 33-37oC) and for radiometric data from 20000 to 21392 units.
Each rectangular area produced a maximum, minimum and average value. All data within
the ROI was saved to file (Excel, Microsoft Corporation, Redmond, WA 98052-6399).
ii) Thermal territory and edge enhancement
Qualitative mapping: To enhance the visual image of the scar line a series of adjustments to
the camera thermal images were made for all participants (greyscale). The first step was to
obtain a reference set of pixel values (oC) for healthy skin distant from the ROI; the umbilical
region was selected. Subtraction of healthy skin temperature from all pixels in the ROI
yielded a new thermogram. To avoid 'negative' values each pixel value was squared. Then,
by calculating the square root for all the values, a new thermal image was generated. By
taking a second square root calculation of the new pixel values, a simple algorithm allowed
scar 'territory' to be highlighted and distinguished from the adjacent healthy skin (Fig 2a).
Data analysis and conversion between pixel values and greyscale thermal image was
achieved using OriginPro 2015. (OriginPro, Origin Lab Corporation, Northampton, MA,
USA). This method is robust, in the sense that, there is no requirement to adjust the colour
pallet or greyscale levels in order to achieve maximum contrast.
Informed by the qualitative edge-detection analysis (Fig 2a) it was possible to proceed to
quantitative analysis. Here a temperature profile line was 'drawn' through the middle of the
scar or infected wound (of the unadjusted grey scale image, Fig 2b) and, with an additional
skin temperature reference line taken at a point above the wound (Fig 2b), a delta plot (Fig
2c) for healthy skin reference minus wound temperature could be constructed. In this way
5
individual ''profile deviations'' would become apparent from the group profile characteristics
(Fig 2C).
iii). Qualitative and Quantitative segmentation of thermal territories
From the grey scale thermogram edge detection process, it became apparent that the
surgical scar or, in patients with confirmed wound infection, both wound and surrounding
regions were not at uniform temperatures. Further analysis of the data for each participant
using the principle of HCS 21 was applied. Briefly, HCS generates a hierarchy of segmented
images by partitioning an image into its constituent regions at levels of 'allowable'
dissimilarity between the different regions. Radiation intensity units were used to produce
HCS pixel clusters. The methodological process is outlined in Fig 3. The pixel cluster
boundaries are plotted with a colour hue comparable to FLIR rainbow HC palette; darker
shades representing regions of low thermal intensity and bright (white, red) pixels indicating
higher thermal intensity. Pixel clusters in regions of uniform temperature will have more
pixels and in regions of high variability in radiation intensity (and temperature) the clusters
will have very minimal number of pixels. For example, In Fig 3B darkest areas (on black and
white thermal image) represent the coldest regions along the scar; the 'cold spots'. Here
variability in radiation intensity is so high that the clusters have barely a couple of pixels of
similar values Fig 3 B. In contrast since the region above the scar line has more uniform
temperature the clusters have more pixels (Figure 3 C). The boundaries (cut-off) for
‘’similarity’’ are determined at a specific level of allowable dissimilarity for clustering; here we
have applied ''allowable dissimilarity'' for thermal profiling of between 10-15%.
In processing the thermal images the HCS helps the user to:
(i) quantify the thermal properties of the different dissimilar regions i.e. between the scar or
wound and the adjacent abdominal thermal territories.
(ii) visualise variability in thermal regions along and amongst the scar ''line'' or wound region
Statistical Analysis
Descriptive statistics of thermal values and comparisons between mean values (independent
student t-test) was undertaken using SPSS (Statistical Package for Social Sciences) IBM
SPSS V22.
Participant Narratives
After thermal imaging had been completed, face-to-face semi-structured interviews were
undertaken with women to determine the acceptability of the procedure to them. Participants
were also asked for their perspectives on the thermal imaging technique and study
processes for future study development. Field notes were obtained during the interviews.
Qualitative data was analysed using content analysis, where the responses given were
grouped into similar themes.
RESULTS
Participants’ Characteristics
Twenty women, predominately of white British origin, aged 20-39 (median 33) years were
recruited (Table 1). After birth, absorbable vicryl/monocryl sutures were used to close the
6
incision (n=8) and non-absorbable Prolene sutures with beads in 11 women. In one woman,
the skin incision was closed with staples on request. The wound was covered during the first
24-48 hours with a dry dressing. Dressings were removed 30 min before imaging
commenced. With the exception of two women (participants 6 and 16) all were admitted for
elective caesarean section and were studied on day 2 post-operatively (Table 1).
Participants 6 and 16 had both been admitted for emergency caesarean section, 11 days
(participant 6) and 15 days (participant 16) previously and had returned to the ward for care
of infected caesarean section wounds. Both had a BMI>25 kg/m2 and both were in-patients
at the time of the study.
Thermal imaging and study data was thus acquired from 18 women at day 2 (and with
apparently ''healthy'' wounds) and two women with a confirmed wound infection (Table1).
Ten of 20 women had a history of previous caesarean section. Two women (who had not
previously undergone caesarean section) gave birth to twins (Table 1). In this series, eight
women were obese (including the two women who returned to hospital with infected
wounds), six women were overweight and six women had a normal booking BMI.
Ambient conditions in the post-natal wards ranged from of 20.8oC to 26.6oC (median 24oC)
with relative humidity; 41 to 73% (median 52%) in still air (<0.02.sec-1). All of the women
were apyrexial at the time of thermal imaging; temporal artery temperatures between 36.2oC
to 37.3oC (median 36.9oC) (Table 1).
Edge analysis and quantitative temperature profiling
Temperature difference profiling is shown for 18 of 20 women (Fig 2C). Data for patient 20
(wound closed by staples) has been omitted. The profile for participant 1 is partly occluded
by adjacent structures (hair). From the difference between healthy ''reference'' skin and
scar/wound temperature (deltaT), pixel differences along the scar/wound margin (Fig 2c)
differ, for the majority of women, by not more than 1oC. In four participants (6, 9, 11, 16)
deltaT reveals four predominately negative deltaT plots, indicating lower scar/wound
temperatures of up to 4oC. In two women, (participant 6, 16) the scar at days 15 and 11
respectively were infected. In participant 11, the scar at day 2 was ''lumpy''; the low deltaT
along the scar indicating that the scar edge was at a consistently lower temperature than
adjacent healthy (reference) skin. Also noted in Fig 2C are intervals of a 2oC deltaT for
participants 9 (see also Fig 1). At this early time-point (day 2) the wound was not considered
to be infected.
HCS isotherm boundaries and regions
Qualitative
In exploring HCS for thermal characteristics of the scar/wound encompassed by the ROI
(see Fig 3A-C) multiple clusters and boundaries were evident on thermal mapping. With the
grey scale FLIR palette providing the visual focus for the scar position on the lower abdomen
(Fig 4), HCS isotherm boundaries were produced to show the variability of the thermogram
for six women where different isotherm boundary maps can be characterised. HCS reveals
the wide variability in thermal clusters that can be distinguished but which are not evident
from the corresponding greyscale palettes (Fig 4).
7
Qualitative Review
1. Scar and adjacent skin clusters show an ROI where the scar and surrounding regions
are virtually indistinguishable within a relatively uniform ('warm') thermal map (participant
18, OV).
2. Small 'islands' of slightly lower thermal values in scar are distinguishable within 'cool'
adjacent thermal territories (participant 5, NORM)
3. Large clusters of lower thermal values are distinguishable within a 'warm' ROI (participant
6- wound infection day 11, OB)
4. Scar profile has higher thermal values than adjacent and surrounding regions (participant
7, OV)
5. The scar profile is indistinguishable on the black and white image and on HCS. The
thermal profile of the scar merges with adjacent and surrounding (cool) skin (participant 8,
OB).
6. The scar profile is evident on HCS and surrounded by three clear regions of low thermal
values which visibly correspond to large areas where exudate has accumulated and
surrounds a large area of denuded skin (participant 16-wound infection day 11).
To further categorise the thermal images, the infected wounds (at later times after surgery)
have prominent ''cold spots' which are identifiable as a marked deltaT on scar boundary
profiling (Fig 2C) and as notable clusters within the ROI commensurate with low thermal
intensity on the colour palette (Fig 4). Of note is participant 9 (Fig 1) where the pattern
features are similar to that observed in wound infection (scar line 'cold spots' within the ROI
are noted at day 2- infection confirmed, day 9).
Quantitative Analysis
On temperature profiling, thermal intensity differences between reference pixel clusters and
the pixel clusters of the scar/wound (see Fig 3 for sites selected) vary (Table 2). Four
participants had large (approximately >2oC) (negative) differences (Table 2) and these
differences in thermal intensity correspond to the ''cold spots'' on FLIR images (Fig 1 A,B)
and HCS (Fig 4). Three of the four participants with a large deltaT had a confirmed wound
infection when re-admitted at day 11 (participant 6) and day 15 (participant 16). For
participant 9, 'cold spots' identified early via IRTI at day 2, subsequently resulted in wound
infection at home on day 9; breakdown of tissue occurring at the original 'cold spot' sites.
Wound swab reports confirmed Pseudomonas species in the wound and with systemic
symptoms of fever and malaise. For the fourth participant (11) with a large, negative thermal
profile, follow up attempts proved unsuccessful. The eventual outcome of the wound in this
participant could not be confirmed. For participants 6, 16, 9,11 the maximum negative
difference between reference and scar for those women with and without
confirmed/suspected SSI achieved significance p=0.006.
If the HCS pixel clusters are examined along and amongst the scar/wound (Table 3), once
again the infected wounds of patients 6 and 16 and the scar region of participant 9 (later
infected) show that large differences occur along the wound i.e. differences in each of the
HCS pixel clusters in this discrete scar/wound area reach 3.5-4.9oC. For women with small
average differences between HCS pixel clusters along the scar, the scar region is virtually
8
indistinguishable from the adjacent abdominal thermal territory (Table 3) as illustrated for
participant 18 (see Fig 4, panel B) .
Narratives
Thematic content analysis of women's perspectives on the use of thermal imaging revealed
four themes; recruitment, personal experience, perceived benefits and repeated imaging.
Recruitment
Women gave varied reasons for taking part in the study with the most common being an
awareness of the need for medical research; with the specific research objective appealing
to many of the women seen as worthwhile and providing possible future benefits. Respect
for research and willingness to help in generating new knowledge were particularly important
factors for women volunteering to participate in the study:
"Without research things cannot progress, so that's why I did it" [P002]
"It's a really good idea to detect whether infections are likely or not as they can be nasty"
[P003]
"It's a study that my children could need the results for [P004]
The second main reason for taking part was a generalised desire to help others:
"I thought it was good to help someone else" [P006]
Women stated that they were mainly recruited at the preoperative clinic, receiving an
information leaflet. Some also expressed their satisfaction at receiving full explanations from
the doctors which impacted on their decision to take part in the study.
Women's views were sought about possible changes to recruitment and study processes for
future trial development. Women had many ideas about how to improve study recruitment
including having posters about the study and wound infection in the pre-operative clinic and
in the unit. They suggested providing leaflets early on in pregnancy or when booking-in to
have an elective Caesarean at about 36 weeks gestation as other information is provided at
this point too. This would reassure women that the study will not involve time away from their
baby. Two women would also have liked midwives to have been involved in study
recruitment, not just doctors:
"Have something about the study while you wait (in pre-op clinic), a poster with some
images of how it would look with a haematoma, normal, with an infection" [P005]
"I think get the midwife involved." [P019]
Personal experience of thermal imaging
When asked about their personal experience of the thermal imaging procedure the majority
of the women described it as "fine", "good" or having "no issues" with it (n=19). One woman
felt the procedure will be helpful, but didn't comment on her experience of undergoing
thermal imaging. If thermal imaging was found to be effective and therefore offered to all
women who had undergone a caesarean section; all of the participants said that they would
accept it:
"I'd be fine with that - with the midwife doing it as part of care" [P017]
"I think it's good - to make it part of routine (care) after having a baby. It can’t make anything
worse having it done." [P003]
9
Description:Dr Mahbubur Rob Siraj, ST7 Obstetrics & Gynaecology, Jessop Wing, between healthy skin reference and wound site, exceed 2oC in women with